nixtla
tsfeatures
nixtla | tsfeatures | |
---|---|---|
10 | 5 | |
1,680 | 332 | |
22.5% | 5.1% | |
9.5 | 5.0 | |
9 days ago | about 1 month ago | |
Jupyter Notebook | Python | |
GNU General Public License v3.0 or later | Apache License 2.0 |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
nixtla
- TimeGPT: Production Ready Time Series Foundation Model for Forecasting
-
Chronos: Learning the Language of Time Series
I do not have a horse in the race, but it is interesting to see open source comparisons to traditional timeseries strategies: https://github.com/Nixtla/nixtla/tree/main/experiments/amazo...
In general, the M-Competitions (https://forecasters.org/resources/time-series-data/), the olympics of timeseries forecasting, have proven frustrating for ML methods... linear models do shockingly well and the ML models that have won, generally seem to be variants of older tree-based methods (ie. LightGBM is a favorite).
Will be interesting to see whether the Transformer architecture ends up making real progress here.
-
[P] Beware of false (FB-)Prophets: Introducing the fastest implementation of auto ARIMA [ever].
Yes, for example we have this paper in long-horizon settings using our library NeuralForecast and this experiment with other of our libraries MLForecast, both of them outperforming autoarima.
-
[P] Deep Learning for time series forecasting (neuralforecast, python package)
Here we did some comparison with prophet in the zillow real-state dataset https://github.com/Nixtla/nixtla/tree/main/utils/experiments/zillow-prophet
- Is linear regression better than prophet? Zillow benchmark
- Prophet vs. Linear Regression on Real Estate: The Zillow Case
-
Automated Time Series Processing and Forecasting
Users can deploy the pipeline in their cloud quickly. We use terraform (https://github.com/Nixtla/nixtla/tree/main/iac/terraform/aws), so it is very simple to deploy the pipeline on AWS. We are working to release versions of terraform on other clouds such as Azure and Google Cloud.
tsfeatures
- tsfeatures: NEW Data - star count:212.0
-
[P] Deep Learning for time series forecasting (neuralforecast, python package)
GluonTS Differences: -GluonTS is written in mxnet, which reduces its adoption. In contrast, NeuralForecast is written in PyTorch. -Including new models in GluonTS tends to be challenging because mxnet 's and the library structure's learning curve are steep. PyTorch-Forecasting Differences: -NeuralForecast hosts some models from our research, including N-HiTS and Transformer-based (Autoformer, Informer, Transformer, etc.) methods specialized in long-horizon forecasting (https://arxiv.org/abs/2201.12886). -And the exogenous variables extension of N-BEATS, the NBEATSx (https://arxiv.org/abs/2104.05522). Extra Features: -NeuralForecast has a wide range of curated datasets used in research to develop and test new models, such as Tourism, M3, M4, M5, EPF, ILI, Traffic, Weather, etc. -NeuralForecast models include reasonable hyperparameter spaces to speed up hyperparameter search, based on our experience. -We include an experiment module that makes it easy to put the entire time series forecasting pipeline into production. -Finally, NeuralForecast is part of a larger ecosystem of time-series analysis and forecasting that includes feature creation (tsfeatures, https://github.com/Nixtla/tsfeatures), machine learning models (mlforecast, https://github.com/Nixtla/mlforecast) and statistical models (statsforecast, https://github.com/Nixtla/statsforecast).
-
Automated Time Series Processing and Forecasting
Thanks for your comments.
We agree that in most cases prophet is not a good benchmark; however, we wanted to use it because it is one of the most used libraries in forecasting. For that reason, we also tested the solution against AWS Forecast obtaining better results.
Besides the better performance and scalability, the pipeline we created considering all the stages of time series forecasting: preprocessing (e.g. missing value imputation), creation of static and dynamic features, forecast generation, and finally evaluation using data sets of important competencies. (https://github.com/Nixtla/tsfeatures)
On the deployment side, the entire pipeline can be quickly deployed in the user's cloud using terraform. This allows for less development time. (https://github.com/Nixtla/nixtla)
What are some alternatives?
darts - A python library for user-friendly forecasting and anomaly detection on time series.
statsforecast - Lightning ⚡️ fast forecasting with statistical and econometric models.
mlforecast - Scalable machine 🤖 learning for time series forecasting.
neuralforecast - Scalable and user friendly neural :brain: forecasting algorithms.
pytorch-forecasting - Time series forecasting with PyTorch
tsai - Time series Timeseries Deep Learning Machine Learning Python Pytorch fastai | State-of-the-art Deep Learning library for Time Series and Sequences in Pytorch / fastai
nixtlats - Deep Learning for Time Series Forecasting.
not-autotools - A collection of awesome and self-documented m4 macros for GNU Autotools